Alipay Investment AI Chatbot
Enhance discoverability and relevance of robo-advisory experience to grow user base
Type
Internship
Team
Intern designer (me) +
1 designer + 1 PM +
4 developer
Timeframe
May - Aug 2021
Shipped in Aug 2021
Tools
Sketch, Photoshop, Illustrator
Context
Alipay Investment AI chatbot (”Zhixiaobao”) is a robo-advisory service that delivers fund recommendation, market insights, and trading behavior analysis to users.
Impact
My design was launched in August 2021, resulting an increased CTR by 7.2% and improved engagement rate by 37.5% in the first month.
New entry point and redesigned in-chat experience (see shipped version in Mandarin).
Problem
Poor engagement rate and stagnant user growth
With the background of slow increase in user base, the chatbot team has also found that the first-time users who entered the chatbot were highly likely dropping off before taking any actions after first round greeting.
NOTE: To comply with the NDA, specific confidential details have been excluded in this case study. Please feel free to reach out and I'd be more than happy to chat about it privately.
Solution
Redesigned in-chat flows and entry points
Feature 1 - New contextualized entry points
Two new entry points on Product Page and Campaign Page enhanced discoverability and provided contextualized prompts.
Entry point on Product Page
Feature 2 - Redesigned first-time chat experience from new entry points
Distinguished different types of actionables, new card components, and new in-chat flow and IA avoided informational confusion, improved confidence, and enhanced trust.
New card component for fund product
Project Goals
Grow active user base by attracting novice investors
By partnering with Conservative Product Team, as the designer from Chatbot Team, I was responsible for adapting first-time conversation flow to attract users and to deliver their low-risk product to them.
Current first-time flow: from main entry point to in-chat product analysis
Metric - How should I measure success?
The current measure of conversion rate (complete buying) might not be applicable to assess this product because beginner investors don’t have proficient knowledge to make trading decisions, and it also contradicts the long-term goal of the product. So I advocated for tracking only two metrics and reached agreement within the team.
Change of success metric
Business goal 1 - Help novice investors be more willing to enter into the service
When the service was pushed to novice investors through entry points, they mostly ignored it.
Business goal 2 - Get novice investors engage more rounds of conversation in chatbot
The very few novice investors that entered into the chatbot didn't even interact once before leaving, which would be the metric of active user.
Major loss of users happens before click and after first greeting
User goal - Help novice investors become more informed and confident about investing
Through the following user study, I concluded this user goal. But how?
Understanding users
Current service focusing on experienced investor doesn’t cater novice investor
Target audience - Why novice investors?
First, the business strategy was to deliver low-risk products which would most likely attract novice investors. Second, the chatbot provided the shortest and easiest path to learn fund products with personalized comprehensibility level from which novice investors can benefit the most.
Pain points - Discoverability, relevancy, comprehensibility
By engaging with a focus group interview on 9 users (including first-time and active), I was able to gather and analyze qualitative data with researchers.
Quotes from interview
Then I determined needs according to findings and their corresponding design goals.
User need 1 - Discoverability
Users want to receive the service with minimal effort when they need it.
Design goal 1 - Move frontline closer
Make sure users have option to enter into service when browsing products.
User need 2 - Relevancy
Users want to know what to expect in the service.
Design goal 2 - Always keep in context
Provide users with contextualized information throughout the entire service session.
User need 3 - Comprehensibility
Users want to utilize the service in a way they can understand.
Design goal 3 - Lower the floor of knowledge
Bring sufficient, understandable and wanted information to users.
Feature 1
Add new contextualized entry points
Problem - Current entry prompts lack sufficient context and therefore attractiveness
Contextualized entry points could provide instantaneity, relevancy, and curiosity.
Current one-line prompt for main entry point is not attractive
Exploration - Direct entries on Product Page
Floating Button: Dissimisiblity and sticky position give users most control, however, it only supports one-line prompts and might block important information when scrolling.
Top Banner: Top location effectively draws attention with highest visual hierarchy, while I eventually let this idea go to because it interrupts user's natural overview-to-chart reading flow and also compromises reachability.
Left: Entry point as floating button. Right: Entry point as top banner
Final design - Contextualized and relevant prompts to improve attractiveness
As a compromise for thumb reachability, avoiding reading flow interruption, and business requirement from other stakeholders, I decided to design the entry as a Bottom Banner as the final deliverable, and variants to accommodate different contexts shown on Product Page.
Final deliverable of new entry point
Feature 2
Redesign in-chat flow and information architecture
Problem - Users’ current mental model of seeing chatbot only as advertisement that pushes them to buy funds
The reasons why they didn’t see it as an educational advisory service were:
Insufficient information affected trust-building.
Chatbot script never informed them it can do more.
They don’t want to type to see detailed analysis (or they don’t know the input field can be autofilled).
Current first-time greeting
Then I determined needs according to findings and their corresponding design goals.
Exploration - What in-chat flow can cater novice investors
After a design-sprinting three concepts, I finally arrived at a combination of enriched dialogue and on-card action button because:
It can give users an overview to stimulate interest.
It can show users its capability and core function as an advisor.
It can separate buying actions from advisory actions which eliminates confusion and encourages further non-buying engagement.
From left to right: advisory CTA as floating button, in-card buying CTA, enriched dialogue
Final in-chat flow (see shipped version in Mandarin)
Challenge - Build a scalable and clear card system
Card system
To make Product Card easily configurable by PMs and Ops and to ensure product info being presented in a comprehensible way for novice investors, I differentiated product info from contextual info. Product info is essential details of a product, while contextual info is configurable, flexible, and subject to change to accommodate contextualized and personalized conversational experience which links external activities to in-chat service. For instance, if user browsed Product Page with star fund manager mentioned, they will see the related information about that fund manage in contextual info section.
Card components breakdown
Evaluation
Cross-functional stakeholder review
Conduct - Concept and user testing with limited resources and time
Because the schedule was extremely tight, I was only able to combine concept and user testing together to validate idea and gather feedbacks, and users preferred the same flow as I assumed and were satisfied with new design except still being confused about specific UX writing.
My super tight schedule (5.5 weeks out of 12-week internship)
Achieve - Legal compliance
UX writing was the key to not just keep users well-informed but also to meet legal compliance for our financial advisory product, so I worked closely with the legal team on this issue.
Approved and rejected terms
Achieve - Compatibility for entire fund product inventory
After discussing with engineers and PMs, I delivered a Product Card library by implementing Ant Design System that covered all product variants (i.e. different types of info) and technical edge cases (i.e. string overflow, missing main info) to accommodate the growing product inventory as well as stay compatible with legacy products.
Variants and edge cases
Results
Business goal achieved
After 5 weeks of design and development, my work was launched in August 2021, and we received positive results.
Reflection
An insightful journey
I was really honored to work with an amazing team. While not everything went as I expected, it was a great journey to explore the complexity in the real-world design process.
Product Strategy
Going into the project, I didn't have a deep enough understanding of the product team's vision and strategy. As a result, some of my design solutions didn't fully align with their long-term goals. Next time, I will make sure to have thorough discussions about product strategy with stakeholders early in the process, so my designs support their strategic objectives.
Documentation
Going into the project, I didn't have a deep enough understanding of the product team's vision and strategy. As a result, some of my design solutions didn't fully align with their long-term goals. Next time, I will make sure to have thorough discussions about product strategy with stakeholders early in the process, so my designs support their strategic objectives.
Scalability and adaptability
Going into the project, I didn't have a deep enough understanding of the product team's vision and strategy. As a result, some of my design solutions didn't fully align with their long-term goals. Next time, I will make sure to have thorough discussions about product strategy with stakeholders early in the process, so my designs support their strategic objectives.
Me and my amazing team
thanks for stopping by;)
Handcrafted with
